Skip to content

bereach-ai/linkedin-prospecting-skills / skills/linkedin-outreach-method

linkedin-outreach-method

Verified

Runs the whole LinkedIn outreach loop: find, sort by persona, voice, one template per persona, send, measure acceptance and replies at 7 days. Start here for any LinkedIn outreach.

FileKindSize
SKILL.mdskill md10.0 KB
references/advanced.mdmarkdown1.9 KB
references/artifacts.mdmarkdown19.9 KB
references/conversation-review.mdmarkdown4.7 KB
references/frameworks.mdmarkdown21.0 KB
references/icp.mdmarkdown20.0 KB
references/measuring.mdmarkdown10.2 KB
references/search-doors.mdmarkdown3.7 KB

references/frameworks.md

View on GitHub
# Reference frameworks behind the outreach loop

How to use this library:
1. No framework here is proven better than another. What is measured is the parts: a first message about them, about their problem, short, with one easy ask. A framework only orders those parts. "Nothing proven" never keeps a slot that fails the expert shape.
2. The baseline is the expert shape in the user's voice (SKILL.md step 4). Frameworks upgrade every failing slot of the user's messages, and each change names the principle it applies; a test compares two versions that both pass.
3. At each step, read its row in Step to framework below and name in the page's grey band only what the page visibly applies (`references/artifacts.md`).

Evidence labels: **Measured** (a study or dataset with a stated method; a study by a publisher that sells the thing measured is a Vendor claim), **Survey** (people answered about themselves), **Practice** (a named practitioner's method, nobody measured it), **Statistics** (a proven mathematical result, no field data needed), **Vendor claim** (a seller's number with no method), **Untraced** (repeated, no primary source found).

Example seller everywhere below (fictional): field calibration software for water networks; persona "network manager".

## Summary

| Framework | Author | Feeds | Evidence |
|---|---|---|---|
| Predictable Revenue | Aaron Ross, Marylou Tyler (2011) | 1 search, 6 results | Practice |
| Two-stage segmentation | Wind, Cardozo (1974) | 1 search, 1 personas | Practice |
| Trigger events | Craig Elias | 1 search, 5 drafts | Vendor claim |
| Social selling, SSI | LinkedIn (2014) | 1 search (warm lists) | Vendor claim |
| The 95:5 rule | John Dawes (Ehrenberg-Bass, for the LinkedIn B2B Institute, 2021) | 4 templates (the ask) | Practice |
| Conversation review | common sales practice | 2 replies, 1 personas, 3 voice | Practice |
| Command of the Message | Force Management | 1 personas | Practice |
| Gap Selling | Jim Keenan (2018) | 1 personas, 4 templates | Practice |
| SPIN Selling | Neil Rackham (1988) | 4 templates, 5 replies | Measured (calls) |
| The Challenger Sale | Dixon, Adamson (2011) | 4 templates | Survey |
| Sandler | David Sandler (from 1967) | 4 ask, 5 replies | Practice |
| NEPQ | Jeremy Miner (7th Level) | 4 ask | Practice |
| Poke the bear, detach | Josh Braun | 4 templates, 5 follow-ups | Practice |
| Multi-touch cadence | Gong (85M cold emails); Expandi | 5 follow-ups | Vendor claim |
| BANT | IBM (date disputed) | background (replies) | Untraced |
| MEDDIC, MEDDICC, MEDDPICC | PTC (1996); Andy Whyte (2020) | background (replies) | Practice |
| SPICED | Winning by Design | 1 personas, background (replies) | Practice |
| Bowtie, volume, conversion, time | Winning by Design (2023) | 6 results | Practice |
| Controlled experiments, one change | Kohavi, Tang, Xu (2020) | 5 halves, 6 results, 7 next test | Statistics |
| Fixed size, no peeking | Evan Miller | 5 halves, 6 results | Statistics |
| Same-age cohorts (7 days) | cohort analysis; Botdog, Belkins timing data | 6 results | Statistics; timing: Vendor claim |
| Simpson's paradox | Edward Simpson (1951) | 1 personas, 6 results | Statistics |

## Step to framework

| Step | Frameworks |
|---|---|
| 1 Plan and personas | Predictable Revenue (ICP first, spears), Two-stage segmentation, Trigger events, Social selling; ICP for search: `references/icp.md`; personas: Command of the Message (before scenario), Gap Selling (problem chart), SPICED (critical event), Simpson's paradox |
| 2 What got replies | Conversation review |
| 3 Voice | Conversation review (their own answered words) |
| 4 Templates | the linkedin-messages skill's `references/frameworks.md`, one framework per message type; the cards below are background |
| 5 Drafts, send, follow-ups | Trigger events (signal line), Multi-touch cadence, Braun's detach (breakup), Kohavi (random halves) |
| 6 Results | Same-age cohorts, Miller (fixed size), Kohavi, Simpson's paradox, Bowtie, Predictable Revenue (measure each stage) |
| 7 Next test | Kohavi (one change at a time, repeat a win), Miller (size before looking) |

## Process and the ICP

### Predictable Revenue (Aaron Ross and Marylou Tyler, *Predictable Revenue*, 2011)
- What it says: write the ideal customer profile before any prospecting; specialise roles (prospecting apart from closing); leads come as seeds (referrals), nets (inbound marketing) and spears (targeted outbound); track the numbers of each stage.
- On LinkedIn: step 1 is a spear: Plan and personas starts from a written ICP and the roles inside it, and splits the people already in the account against it before any search; step 6 reads each stage on its own (accepted, then answered).
- Example: "ICP: water utilities serving 20,000 to 200,000 people, France; role inside: network manager."
- Use when / not when: always for the order (ICP, then list); not for its response benchmarks.
- Evidence: Practice; its benchmarks are second-hand retellings.

### Two-stage segmentation (Yoram Wind and Richard Cardozo, "Industrial Market Segmentation", 1974)
- What it says: segment the organisations first (macro), then the people inside them (micro).
- On LinkedIn: the ICP picks the companies at step 1, the persona cards pick the people and what to say, both at step 1. Two stages, in order.
- Example: utilities of a given size first, then "network manager" against "operations director" inside them.
- Use when / not when: whenever a list mixes roles; not to sort leads by personality.
- Evidence: Practice; a conceptual model, never tested on outreach.

### Trigger events, the window of dissatisfaction (Craig Elias, "Seeing through the window of discontent", CustomerThink)
- What it says: write when something changed for them (new role, hiring, a project, a tool change), while the current way has just stopped working.
- On LinkedIn: step 1 slice of signal leads (`signal:<name>`), invited first; step 5 the signal is the lead's first line in both halves.
- Example: "Saw the utility is hiring two leak detection technicians this month."
- Use when / not when: when the event is theirs, recent and tied to a problem; not a funding round named with no "so what".
- Evidence: Vendor claim; activity premises measured by vendors only, Elias's "74%" untraced.

### Social selling and the Social Selling Index (LinkedIn, 2014)
- What it says: build a professional brand, find the right people, engage with insights, build relationships (four pillars of 25 points).
- On LinkedIn: a warm list (people who reacted to or commented on a post on the topic) is a step 1 source with a real premise per person; the score itself is never a goal.
- Example: "Saw your comment on the post about meter drift after winter."
- Use when / not when: when a real tie exists; not mechanical warm-ups (a visit, a like), which nobody measured.
- Evidence: Vendor claim; LinkedIn's own uplift figures, no method.

### The 95:5 rule (John Dawes, Ehrenberg-Bass Institute, for the LinkedIn B2B Institute, 2021; made known by Peter Weinberg and Jon Lombardo)
- What it says: at any moment only about 5% of B2B buyers are in market.
- On LinkedIn: step 4 the ask must be worth answering for someone NOT buying this quarter: an interest question or an offer they can keep, never a demo.
- Example: "Want the two-page checklist we use before a summer pressure campaign?"
- Use when / not when: every first message ask; not as a precise number.
- Evidence: Practice; purchase-cycle arithmetic, given as an estimate.

## Who, and what they struggle with

### Conversation review (common sales practice, no single author)
- What it says: before writing, read your own past conversations, count which openers and asks got answers, note the prospects' exact words.
- On LinkedIn: What got replies (step 2, once): interested answers against sent by opener and ask, politeness apart; Keep, Drop, Their words feed the persona cards, the voice samples and the offer. Counts are hints for the slot judgment, never a template kept (`measuring.md`, Past messages).
- Example: "4 of 6 openers on a pressure incident got an interested answer; 0 of 5 with a demo ask: a hint, not proof."
- Use when / not when: when the user has written first to prospects before; whatever the count, a hint, never a verdict.
- Evidence: Practice; small, self-selected samples.

### Command of the Message (Force Management)
- What it says: know the buyer's before scenario, its negative consequences, the after scenario, the business outcomes, the required capabilities and the metrics.
- On LinkedIn: the before scenario and its cost become the persona card's work moment line; each message draws one line from it.
- Example: "Calibrating meters means a technician on site for every sensor, so half the network is checked once a year."
- Use when / not when: writing a persona card; not recited whole in a message.
- Evidence: Practice; component lists disagree between summaries.

### Gap Selling (Jim Keenan, *Gap Selling*, 2018)
- What it says: buyers buy change: map the current state (problem, impact, root cause), the future state and the gap; build a problem identification chart before outreach.
- On LinkedIn: the chart is the list of problems a persona card and a template may name; an observable current state (hiring, a new role) can open the message.
- Example: "Most network managers we talk to find out a sensor drifted when the leak report disagrees."
- Use when / not when: building the cards and choosing a template's problem; not a gap speech in a cold message.
- Evidence: Practice.

### SPICED (Winning by Design)
- What it says: Situation, Pain, Impact, Critical Event, Decision; the critical event is a dated reason it must happen now.
- On LinkedIn: the card's Why now is a critical event; in a reply, the next question follows SPICED order, one at a time.
- Example: "Is the regulator's audit in spring the date you are working to?"
- Use when / not when: replies and Why now lines; not five questions in one message.
- Evidence: Practice; vendor pages and summaries only.

## The message

### SPIN Selling (Neil Rackham, Huthwaite, *SPIN Selling*, 1988)
- What it says: Situation, Problem, Implication, Need-payoff questions; in large sales, closing techniques lower success and advantages count less than benefits; what works in small sales does not carry to large ones.
- On LinkedIn: no Situation question (read the profile instead); one Problem or Implication line in the first message; Need-payoff only in a reply, once they said the problem is real.
- Example: "When a pressure sensor drifts, how long before the night-flow figures stop meaning anything?"
- Use when / not when: the problem line of every template; never a closing line in message 1.
- Evidence: Measured; about 35,000 calls observed by the author's own firm, data unpublished; meetings, not messages.

### The Challenger Sale (Matthew Dixon and Brent Adamson, CEB, 2011; *The Challenger Customer*, 2015)
- What it says: teach something new about the buyer's business, tailor it, take control; later: give each person what they can take to their team.
- On LinkedIn: "teach" is one insight they may not have, as a template test arm; tailoring is the persona card; "take control" never goes in a cold message.
- Example: "With lab rotation, a drifted sensor only shows up once it is back from the lab, months after the readings it skewed."
- Use when / not when: when the user has a true, sourced insight; not a lesson on their own trade (check `leçon sur leur métier`).
- Evidence: Survey; managers rated their reps, peer-reviewed critiques exist.

### Sandler (David Sandler, Sandler Training, from 1967)
- What it says: an up-front contract (purpose and outcomes agreed, no included), then a pain funnel from broad to personal.
- On LinkedIn: the ask leaves an explicit easy no; the pain funnel is for replies, one question per message.
- Example: "If it is not on your list this year, just say so and I will leave it there."
- Use when / not when: the ask and replies; never stacked questions in a DM.
- Evidence: Practice; no published outcome data.

### NEPQ (Jeremy Miner, 7th Level)
- What it says: neutral, curious questions; "safe no" asks such as "Would you be opposed to...?".
- On LinkedIn: one wording for the step 4 ask; the stage machinery stays out.
- Example: "Would you be opposed to me sending the one-page version?"
- Use when / not when: to upgrade an ask that fails the shape; never over an ask that passes it.
- Evidence: Practice; outcome claims are the vendor's own.

### Poke the bear and detach (Josh Braun)
- What it says: a neutral question on the cost of the status quo, never a leading one; every ask leaves an easy no; no chasing, no guilt.
- On LinkedIn: the problem line of a template as a neutral question; the last follow-up closes the loop.
- Example: "How do you know which of your 300 sensors drifted since the last campaign?"
- Use when / not when: the body of a first message and the breakup; not as the ask itself (a problem question as the call to action measured worse in email data).
- Evidence: Practice; breakup lift measured on email by vendors.

## Sending and replies

### Multi-touch cadence (Gong, "How to master cold email", 85M cold emails, also presented with 30 Minutes to President's Club; Expandi LinkedIn benchmarks, 2026)
- What it says: on LinkedIn, sequences of three messages replied best and five or more worst; each touch brings something new; a short breakup message lifts late replies.
- On LinkedIn: one first message and two follow-ups per persona (step 4), follow-up 1 only after the day-7 read, an answer stops the sequence for the whole firm.
- Example: follow-up 2: "If calibration is not a topic this year, no problem, I will leave it here."
- Use when / not when: the shape of every sequence; not touch counts or spacing, which nobody measured on LinkedIn.
- Evidence: Vendor claim; 6.7M messages (Expandi), 85M emails (Gong), observational.

### BANT (IBM, mid twentieth century, decade disputed)
- What it says: qualify on Budget, Authority, Need, Timing.
- On LinkedIn: a checklist for what a reply has told you, never questions in a first message; feeds the proposed answer on Follow-ups to send.
- Example: a reply "we redo the network next year" fills Need and Timing; Budget stays unknown.
- Use when / not when: sorting interested replies; not as an opener, and never as a gate on who is invited.
- Evidence: Untraced origin; widely used as Practice.

### MEDDIC, MEDDICC, MEDDPICC (PTC, 1996, credited to Dick Dunkel, Jack Napoli and John McMahon; Andy Whyte, *MEDDICC*, 2020)
- What it says: Metrics, Economic buyer, Decision criteria, Decision process, Identify pain, Champion; later Competition and Paper process.
- On LinkedIn: in a reply, which letter is missing names the next question; "Economic buyer" can point to a second persona at the same firm (held while the first is in conversation).
- Example: "Who besides you signs off on a new tool for the meter fleet?"
- Use when / not when: an interested reply in a complex sale; never in a first message.
- Evidence: Practice; origin credit varies between accounts.

## Measuring and the next test

### Bowtie data model (Winning by Design, 2023)
- What it says: measure every stage of the customer journey on three numbers: volume, conversion, time.
- On LinkedIn: invitations sent (volume), accepted and answered within 7 days (conversion), days to accept and reply (time) on Results.
- Example: "90 invited, 34 accepted at 7 days, 5 answered."
- Use when / not when: the shape of Results; not its benchmarks.
- Evidence: Practice.

### Controlled experiments, one change at a time (Ron Kohavi, Diane Tang, Ya Xu, *Trustworthy Online Controlled Experiments*, 2020)
- What it says: compare a change against a control, random split, one change at a time; an A/A split shows how often chance alone crosses the line (about 1 in 20).
- On LinkedIn: the two halves of a persona (shuffle, alternate), the persona template against the user's shape, one slot changed (`deux changements` blocks); repeat a win once before rebuilding around it.
- Example: same message, only the ask swapped: "Want the checklist?" against "Would the one-page version be useful?".
- Use when / not when: every test; under 30 expected first messages, a baseline only.
- Evidence: Statistics; random split and A/A false win rate.

### Fixed size, no peeking (Evan Miller, "How Not To Run an A/B Test")
- What it says: decide the size before looking; stopping at the first "significant" gap made a null test win 26.1% of the time in his example; people per arm about 16 p(1 - p) / gap squared.
- On LinkedIn: Invitations to send computes the size and the Results date in advance; Results is drawn once, on that date.
- Example: 10% replies hoping for 20% needs about 200 first messages per half (16 x 0.15 x 0.85 / 0.1 squared, p the average rate).
- Use when / not when: sizing and reading every test (big changes only: `measuring.md`).
- Evidence: Statistics; his simulation and the sample size formula.

### Same-age cohorts, the 7-day read (cohort analysis; timing data from Botdog and Belkins)
- What it says: acceptances and replies arrive over days (Botdog, 16,492 requests: "88% of all acceptances happen within one week"; Belkins: 91% within 14 days), so compare periods at the same age.
- On LinkedIn: the `read7` tags and `connections.trend`: what went out, and what of it was accepted or answered within 7 days; younger periods are counting.
- Example: "Batch of 3 Nov read on 10 Nov, never on 5 Nov."
- Use when / not when: every rate on Results; never a count to date.
- Evidence: Statistics (cohort comparison); timing figures: Vendor claim (Botdog 16,492 requests, Belkins 152,636 contacts).

### Simpson's paradox (Edward Simpson, 1951)
- What it says: a version can win inside each group and lose overall when the group mix differs.
- On LinkedIn: templates are compared inside one persona only; never a sum across personas on Results.
- Example: the checklist ask wins with network managers and with directors, yet loses overall because one half held more directors.
- Use when / not when: any comparison across mixed lists; one persona alone needs no adjustment.
- Evidence: Statistics; a mathematical result (Simpson, 1951).

## Further reading

- Aaron Ross and Marylou Tyler, *Predictable Revenue*, 2011.
- Yoram Wind and Richard N. Cardozo, "Industrial Market Segmentation", Industrial Marketing Management 3(3), 1974.
- Neil Rackham, *SPIN Selling*, 1988; Huthwaite interview: https://www.huthwaiteinternational.com/blog/neil-rackham-research-spin
- Matthew Dixon and Brent Adamson, *The Challenger Sale*, 2011; Brent Adamson, Matthew Dixon, Pat Spenner, Nick Toman, *The Challenger Customer*, 2015; HBR, "The End of Solution Sales", 2012: https://hbr.org/2012/07/the-end-of-solution-sales
- Noel Capon, "Revisiting The Challenger Sale: breakthrough built on a flimsy foundation", 2015: https://business.columbia.edu/faculty/research/revisiting-challenger-sale-breakthrough-built-flimsy-foundation
- Jim Keenan, *Gap Selling*, 2018; Gong guide: https://www.gong.io/blog/gap-selling
- Sandler pain funnel, Gong guide: https://www.gong.io/blog/sandler-pain-funnel/
- Force Management, Command of the Message: https://forcemanagement.com/blog/how-we-describe-command-of-the-message-to-sales-and-marketing-leaders
- Josh Braun, "Ditch the pitch, poke the bear": https://joshbraun.com/ditch-the-pitch-poke-the-bear/ and "Cold email CTAs": https://joshbraun.com/cold-email-ctas/
- Jeremy Miner method, secondary guide: https://www.11x.ai/guides/jeremy-miner-sales-training-method
- Craig Elias, trigger event selling: https://customerthink.com/seeing_through_window_discontent/
- Gong, "How to master cold email" (85M cold emails): https://www.gong.io/resources/guides/how-to-master-cold-email-get-the-data-backed-guide-based-on-85-million-emails ; a version presented by 30 Minutes to President's Club (co-credit unconfirmed): https://tactics.30mpc.com/hubfs/The%20Ultimate%20Cold%20Email%20Data%20Report-1.pdf
- Expandi, "State of LinkedIn Outreach H2 2026": https://expandi.io/state-of-linkedin-outreach-h2-2026/
- LinkedIn Sales blog, "What does a rising Social Selling Index mean", 2014: https://www.linkedin.com/business/sales/blog/modern-selling/what-does-a-rising-social-selling-index-mean-for-sales-teams
- John Dawes, Ehrenberg-Bass Institute, for the LinkedIn B2B Institute, 2021 (the source of the 95:5 rule); made known by Peter Weinberg and Jon Lombardo, "The 95:5 rule", Marketing Week, 2021: https://www.marketingweek.com/peter-weinberg-jon-lombardo-95-5-rule/
- Andy Whyte, *MEDDICC*, 2020.
- Winning by Design, SPICED summary: https://resources.rework.com/libraries/pipeline-management/spiced-framework ; bowtie model, 2023: https://winningbydesign.com/press/winning-by-design-releases-updated-bowtie-model-and-benchmarks-to-provide-go-to-market-teams-with-a-standardized-customer-journey-and-data-model/
- Ron Kohavi, Diane Tang, Ya Xu, *Trustworthy Online Controlled Experiments*, Cambridge University Press, 2020.
- Evan Miller, "How Not To Run an A/B Test": https://www.evanmiller.org/how-not-to-run-an-ab-test.html ; sample size calculator: https://www.evanmiller.org/ab-testing/sample-size.html
- Botdog, LinkedIn acceptance rates: https://www.botdog.co/blog-posts/linkedin-acceptance-rates ; Belkins, LinkedIn response time study: https://belkins.io/blog/linkedin-response-time-study
slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack."Stars: 198kUpdated Claude CodeCodexVerifiedInstall
ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read.Stars: 196kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report.Stars: 196kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt.Stars: 196kUpdated Claude CodeCodexVerifiedInstall